MTH 563

Project task 5

Student C

(Please open in Colab) The HCP dataset comprises resting-state and task-based fMRI from a large sample of human subjects. The subset of the dataset includes time series data that has been preprocessed and spatially-downsampled by aggregating within 360 regions of interest (ROI).

Basic parameters

Downloading data

The rest and task data are shared in different files, but they will unpack into the same directory structure.

They also separately provide some potentially useful behavioral covariate information.

Loading region information

Downloading either dataset will create the regions.npy file, which contains the region name and network assignment for each parcel.

Detailed information about the name used for each region is provided in the Supplement to Glasser et al. 2016.

Information about the network parcellation is provided in Ji et al, 2019.

We also provide the parcellation on the fsaverage5 surface and approximate MNI coordinates of each region, which can be useful for visualization:

Helper functions

Data loading

Task-based analysis

Resting-state analyses

Load a single run of resting-state data:

Load a concatenated resting-state timeseries (using all runs' data) for each subject:

Run a simple correlation-based "functional connectivity" analysis

Generate a correlation matrix (showing "functional connectivity" or FC) for each subject and plot the group average:

Plot the profile of FC values between a particular "seed" parcel and every parcel in the dataset, separated by hemisphere:

Threshold the correlation matrix to produce a connectome, and plot it:

Exploring and analyzing network

Task performance measures

The dataset also includes aggregate behavior for each task run stored in task-specific .csv files. It is possible to load and work with these files using numpy:

we could see that frontoparietal network connectivity significantly predicts 2-back reaction time, independent of default mode and somatomotor network connectivity